Excavating-RoI-Attention-for-Underwater-Object-Detection

This paper was accepted by ICIP2022 (arXiv:2206.12128).

Dependencies

Python == 3.7.16

PyTorch == 1.7.0

MMDetection == 2.20.0

MMCV == 1.4.6

Numpy == 1.21.2

Installation

The basic installation follows with [mmdetection] [document]. It is recommended to use manual installation.

Dataset

We use dataset UTDAC2020, the download link of which is shown as follows.

https://drive.google.com/file/d/1avyB-ht3VxNERHpAwNTuBRFOxiXDMczI/view?usp=sharing

After downloading all datasets, create UTDAC2020 document.

$ cd data
$ mkdir UTDAC2020

It is recommended to symlink the dataset root to $data.

Excavating-RoI-Attention-for-Underwater-Object-Detection
├── data
│   ├── UTDAC2020
│   │   ├── train2017
│   │   ├── val2017
│   │   ├── annotations

This model is also applicable to Pascal VOC and COCO datasets.

COCO: https://cocodataset.org/#download

PASCAL VOC: http://host.robots.ox.ac.uk/pascal/VOC/

Other underwater dataset: https://github.com/mousecpn/Collection-of-Underwater-Object-Detection-Dataset

Train

If you want to use Pascal VOC or COCO dataset, lease change the dataset type under the roitransformer_r50_fpn_1x_coco.py file.

$ python tools/train.py configs/faster_rcnn/roitransformer_r50_fpn_1x_coco.py

Test

$ python tools/test.py configs/faster_rcnn/roitransformer_r50_fpn_1x_coco.py <path/to/checkpoints>

Citation

@inproceedings{liang2022excavating,
  title={Excavating RoI Attention for Underwater Object Detection},
  author={Liang, Xvtao and Song, Pinhao},
  booktitle={2022 IEEE International Conference on Image Processing (ICIP)},
  year={2022},
  organization={IEEE}
}

Acknowledgement

This work is suported by Science and Technology Development Fund of Macau (0008/2019/A1, 0010/2019/AFJ, 0025/2019/AKP).

And thanks MMDetection team for the wonderful open source project!